The Exhaust of Attention
We really do not wish to see The Hinternet degenerate into a never-ending “AI Says the Darnedest Things!” column, and we promise this will be the last of that sort. Such pieces are, in 2025, fast becoming the essay-length equivalent of the inane social-media posts we began seeing some years ago, telling us all the darnedest ways autocorrect was changing our words, from “causal” to “casual”, from “converb” (look it up) to “convert”, and so on. We get it! The machines are systematically warping everything we try to say! It’s so funny! And now they’re giving us all our writing ideas too, because we have no thoughts of our own anymore! Haha!
Still, we think you’re going to want to hear this.
Fellow Substack authors will be familiar with the columns of data that look something like the following:
With e-mail addresses safely eliminated, this list shows the top readers of JSR’s most recent piece. But in what way are they “top” exactly? Their ranking is determined by the number of “opens” attributed to them; but what is an “open”? We have long had trouble believing that any individual reader might ever have reason to open and close and open the same piece 224 times. We have therefore supposed that that figure must be something more like the total number of times the reader has caused the piece to be opened —if only there were a vestigial middle verb for this action, such as we still have in a few exceptional cases in English, for example the lumberjack who fells a tree—, whether by opening it themselves or, for the great majority of opens attributed to them, sharing it in their network so that other people might open it.
Determined to find out whether this was in fact what was happening, we turned, as we now do so naturally and easily, to ChatGPT, which informed us that the high number of “opens” reflects “the number of times the tracking pixel loaded under that subscriber’s ID, due to email client preloading, forwarding, or security scans — and not necessarily repeated manual views by the reader.” Even if we can’t know how many of the individual “opens” in the figure involve an actual human being who is reading and cognizing the words in the piece, the LLM told us, the figure is generally a good “proxy for engagement”. It may be understood, the non-conscious entity explained, as a sort of “attention exhaust” — a “residual trace of someone caring about your work”.
This is curious.
Those of you who were present at the 2016 exhibition of objects from the “W-Cache”, hosted by ESTAR(SER) in collaboration with the Museum of Jurassic Technology in Culver City, California, will perhaps have retained at least a vague mental image of the “Douce Virgula”. To refresh your memory, you may download the entire catalog of the exhibition below. The virgula is the fourth of 33 items, lovingly photographed and described:
This object, named for the English antiquarian Francis Douce (1757-1834), is thought to have functioned as a sort of “dousing rod” for measuring the amount of attention that had previously been bestowed on any material object placed before it — an “applause meter”, as it has sometimes been described, except that what it measures is not so much current decibels, as past applications of attention, most instances of which, likely, made no sound at all.
The Douce Virgula was subsequently described here at The Hinternet in a guest piece from Contributing Editors Cisco T. Laertes and Eigil zu Tage-Ravn in 2021. They write:
Douce, who held an unsuccessful tenure as the “Keeper of Manuscripts” at the British Museum at the start of the nineteenth century, had a strong interest in history, aesthetics, and magic, and was an intimate of the notable English society sculptor Joseph Nollekens (1737-1823), from whom he (Douce) inherited a small fortune. Nollekens was obsessed with his own fame, and the fame of his works, and rumors have long persisted that he sought out the assistance of a Galloway “cunning woman” who assured him she could use dousing rods to determine whether his works were receiving more or less attention than those of his arch-rival, the notable Georgian sculptor John Bacon (1740-1799).
Laertes and zu Tage-Ravn conjecture that Douce likely picked up the notion of attentional dousing from Nollekens. “But whereas the conceited Nollekens seems to have been interested exclusively in using the device pictured above as a kind of silent ‘applause meter’ for his own creations, Douce activated it for quite the opposite purpose: to locate marginal, ignored, or forgotten works of art.”
So then, what was this dousing if not a search for attention exhaust, the residual trace of someone caring about your work? This, it turns out, is something inventors and conjurers of various stripes have long endeavored to measure — even if, for a long time, we did so by means of dubious technologies that belonged more to the realm of divination than of scientific discovery (though some of the more “postmodernist” members of our Editorial Board would deny that there’s much difference — yes, we’re looking at you, JSR). The Great Dematerialization —the shift of a good many of the “objects” of our concern from a physical to a virtual setting— is making the measurement of attention exhaust a matter of precise quantification, rather than of intuition, vibes, or the mystifications of any cunning woman or shaman.
Most of us, however, are still following —vanity of vanities!— in the footsteps of Nollekens, while what the world plainly needs is for more to proceed in a Doucean spirit, Doucement you might say: to approach our digital objects with the same solicitude for the marginal and forgotten as Douce once approached the shelved specimens —the Khoikhoi spiderweb hats, the zoomorphic dung effigies— in the storage rooms of the British Museum.
However we use it, though, this much is clear — we’re getting what we always wanted. And as to whether attention actually gives off “exhaust” or not, just ask yourself: are you not exhausted?
Don’t forget to read and subscribe to
, the Substack wing of the operations of our friends over at the School of Radical Attention. Here at The Hinternet we only write about attention, and its current crisis, now and then, and usually only in connection with other matters of equal interest to us. SoRA is on that beat 24/7!And don’t forget, also, that we are now less than two months away from the deadline for the Inaugural Hinternet Essay Prize Contest. If you intend to submit, we are here to tell you that you better get to work!
Our own JSR gave a lively and engaging interview to Sam Kahn over at Republic of Letters. Read it!
It is a testament to my heat-addled brain that I read the headline as "The Exhaustion of AI".
There's something else quite interesting here, unless I'm being fooled myself by a bit of Hinternet trickery. Did the LLM really use the phrase "attention exhaust?" If so, that's a fascinating construction I don't think I've ever seen before, but it recalls "data exhaust," which Shoshana Zuboff wrote about (I believe she popularized the term a bit). If I'm right that this term hasn't been used before in this way, it's evidence of the LLM "seeing" a relationship between the way "attention" and "data" are talked about in its training data and intuiting that "attention exhaust" is a possible term based on statistical relationships. This is one interesting way to use LLMs--when they "hallucinate" or do something novel, it's not simply a mistake, it's telling us that there's some sort of relationship among different words and ideas that we (humans) might not have noticed but is present in the vast amounts of text that we've produced.